Kenya Long-term Exposure Experiment (KLEE)

Kenya Long-term Exclosure Experiment (KLEE)


KLEE Data Archive


Project Summary

In 1992, we initiated long-term experimental and descriptive studies of community ecology in a dry savanna woodland ecosystem on the Laikipia Plateau in Kenya, at the Mpala Research Centre. The centerpiece of this research is the Kenya Long-term Exclosure Experiment (KLEE), established in 1995. Using different types of exclosures, we are demonstrating the separate and combined effects on community structure of herbivory by cattle, elephants (and giraffes), and other indigenous large herbivores (zebras, gazelles, elands, buffaloes, hartebeests, oryx). This work is being done in collaboration with Karatina University, Egerton University, the University of Nairobi, the National Museums of Kenya, and the Kenya Wildlife service.


Over 70% of Africa is arid or semi-arid, and this land is under continuing pressure for increased grazing and conversion to agriculture, risking disastrous consequences (species extinction, desertification, and the contribution of changing land use to global warming). This long-term study will help clarify both natural and human-induced dynamics in an arid land African ecosystem, in the context of changing land use.


Over 70% of the land area of Africa is arid or semi-arid. Rapidly growing populations and ongoing economic development are putting tremendous pressures on these lands (Bernard et al. 1989). These pressures can result in local, regional, and global problems, including soil erosion, threats to endangered species such as elephants and rhinos, and desertification, with its potential effects on global warming.  In response to these serious problems, the Government of Kenya (through the National Museums and the Kenya Wildlife Service) is collaborating with the Smithsonian Institution and Princeton University to establish a research and training center for arid ecosystem studies. This center is located on Mpala ranch, a 45,000 acre property in the center of the Laikipia Plateau in Kenya. The goal of this center is to provide the information, expertise, and training that will allow Kenya (and other nations in the region) to wisely manage, develop, and conserve their arid land resources.  For virtually all indigenous animal species in Kenya, there are more individuals outside protected areas than inside them, mostly on arid and semi-arid lands. The future of these populations will depend on the interaction between enlightened rangeland management and wildlife needs. There is growing evidence that many wildlife species are not incompatible with intensive livestock production, and can even be beneficial. Also, wildlife can provide much needed additional revenue in the form of tourism or game ranching. A mixed strategy of wildlife and cattle may be optimal, both for economic return and for the maintenance of biodiversity. There is considerable literature on the effects of various cattle grazing and burning strategies on range quality in Africa, and on the effects of indigenous herbivores on the vegetation of protected areas. However, almost little is known about the interactions among wildlife, domestic stock, and vegetation, and there have been no controlled experimental studies of the separate and combined effects of wild and domestic herbivores on vegetation and on each other. This information is vital if we are to make informed decisions about the future of wildlife on the vast unprotected grazing lands of Africa. We are engaged in a long-term experimental study of these interactions at the Mpala Research Centre in Laikipia, a working cattle ranch and an area rich in wildlife. For example, the Laikipia ecosystem has the second largest population of elephants in Kenya; as many as 900 elephants periodically occur on Mpala. There are also at least twenty additional species of large indigenous mammals.

Experimental Design

KLEE SchematicIn 1995, we established a set of exclosure experiments that separately excludes livestock, wild large mammal herbivores in general, and elephants (and giraffes) specifically. A large tract of Acacia savanna woodland of several hundred hectares was selected for this study on Mpala ranch. Preliminary surveys of the study area allowed us to identify suitable sites for the establishment of experiments, to pair the most similar study sites for greater statistical discrimination, and to provide a baseline against which to measure long-term change. In addition, these experimental plots were located to includes part of the long-term glades that provide several crucial ecosystem functions. Eighteen similar four-hectare plots were selected for the following treatments (three replicates of each):

  1. O: Full fencing to exclude all large herbivores.
  2. C: Full fencing, but cattle allowed to graze periodically.
  3. WC: High single-strand electric fencing to exclude elephants (and giraffes), other herbivores allowed.
  4. W: As in #3, but cattle excluded.
  5. MWC: Unfenced, cattle allowed to graze.
  6. MW: Unfenced, cattle are not allowed by herders to graze (low posts as visual barriers).

In addition, nine rodent exclosures were initially placed in the total exclusion plots.

The two types of fencing we are using, for elephants and for other wildlife species, respectively, have a proven track record in Kenya. Dung counts and monitoring of marked twigs have continued to show that the exclosures are 90-100% successful in reducing both the presence of targeted herbivores and their actual herbivory. Ongoing surveys include quantification of floristic and physiognomic vegetation structure and soils analysis for chemical and structural characteristics. Every June, vegetation is surveyed at 20-100 fixed locations within each plot (depending on taxon). Line transects are regularly surveyed for fresh dung of all large vertebrate herbivore species, identified by experienced local scouts, and using a reference collection to be established at the research center.

Additional experimental layers

Intense grazing

The cattle (C) treatments in the KLEE experiment are imposed though multiple cattle "runs" each year, whose overall effect is meant to replicate both the stocking densities and patterns that are common on the privately-managed ranches in Laikipia (i.e., moderately-stocked'). In 2008, we added a 50m x 50m sub-treatment in each of the nine C plots that was meant to more closely replicated the much higher cattle stocking intensities that are common on teh communal pastoral areas.


In 2013, we carried out 18 controlled burns within our KLEE rangeland plots. These fire experiments are revealing complex interactions among rangeland drivers, including clear demonstrations of strong synergy (context-dependence) between fire and herbivory and subsequent trophic cascades involving both livestock and native biodiversity. In 2018, we re-burned these 18 subplots, and carried out 18 more (36 burns in 2018), testing for year effects, multiple-burn effects, and the effects of past burn x herbivore interactions on fire behavior and its consequences.

Summary of results

Kenya Long-term Exclosure Experiment (KLEE), a large-scale factorial experiment investigating the effects of native and domestic herbivores on each other and on their shared savanna landscape . The National Science Foundation (NSF) has supported this project with four five-year LTREB grants, with additional support from The National Geographic Society and the Smithsonian Institution.  We continue to attract other research groups whose work depends on our experimental plots. The overall KLEE project has produced >140 peer-reviewed papers, including papers in Science (x6), PNAS (x3), Ecology (x15), Journal of Ecology (x9), Ecological Applications (x11), Oecologia (x7), and Oikos (x3). More than 100 of these are from the exclosures themselves, making KLEE the most productive field experiment ever carried out on the continent of Africa. The project has also produced 28 outreach publications and supported 28 masters and doctoral dissertations, more than half of which were carried out by Kenyan students. An additional Broader Impact has been our commitment to using the KLEE study to inform livestock management and wildlife conservation in multi-use landscapes, through peer-reviewed publications, outreach publications, and direct stakeholder engagement.

Several of these studies are part of our integration of the KLEE experiments with broader landscape land use analyses. The long-term exclusion of large mammalian herbivores is beginning to result in differences in soil chemistry. In addition, we have demonstrated herbivore impacts on soil cracking and hydrology as well as the dynamics of soil microbes and carbon sequestration.


Trophic Figure

  Some of the many relationships demonstrated in the Kenya Long-term Exclosure Experiment (KLEE). Boxes in bold represent manipulated variables in replicated controlled experiments.

  Our analysis of long-term time series data demonstrated that the most important axis of variation in rangeland plant community structure is driven by overall grazing pressure, regardless of whether this is from cattle or wildlife. Similarly, we used both remote-sensing and field-based assessments to demonstrate that herbivory by cattle and/or wildlife increases ecosystem primary productivity (stimulation of grass growth by herbivory), except at the highest herbivory levels. Livestock and wildlife, however, do differ in their effects on spatial and temporal variability in herbaceous productivity. Most recently, we have begun to use the long-term data to explore changes in community resilience in this system in the face of multiple interacting stressors.

This large-scale, long-term exclusion experiment is demonstrating how different guilds of large mammals are important drivers of community and ecosystem dynamics in a wooded grassland ecosystem in Laikipia, Kenya. From 1996 to 1998, rainfall was at or above average.   One of the effects of this high rainfall was that the productivity of the vegetation was greater than consumption by herbivores.   Differences among herbivore treatments were muted, as were competitive effects among herbivores.   However, the severe 1999-2000 drought accentuated profound differences among the plots, many of which had been underway even during wetter years.   These differences have continued to be evident in subsequent years (2001-2002), with more average rainfall. There have been four major areas of research thus far- competition and compensation (effects of treatments on vegetation and animals), ecology of Acacia drepanolobium and other woody spp., induced plant defense, and habitat heterogeneity. Compensation and other responses of animals to herbivore exclusion Our exclusions of large mammalian herbivores have caused compensatory responses in non-excluded herbivores.   Rodent numbers quickly doubled in the total exclosure plots compared to controls (Keesing 1998), and this difference has continued throughout the study period, despite large fluctuations in the absolute numbers of rodents associated with rainfall variation.   In contrast, compensatory responses of (non-excluded) large mammals were initially muted, perhaps because of unusually high rainfall in 1996-1998.   However, since 1999/2000 strong treatment effects have been evident. Cattle, wildlife, and megaherbivores all significantly reduced grass cover.   However, these effects were not additive.   Cattle fully compensated for the absence of wildlife or megaherbivores, but neither wildlife nor wildlife and megaherbivores together fully compensated in grass cover for the absence of cattle. Wildlife, especially megaherbivores, significantly reduced cover by forbs by up to 35%.

Grass Cover


Forbs Fig

Recently, we have demonstrated complex patterns of competition and compensation between cattle and wildlife.   A) Cattle feeding in plots from which wildlife have been excluded (which have more grass) have significantly higher bite rates and lower step rates than cattle in plots shared with wildlife.   Competition with wildlife also significantly affects the diet of cattle.   Although several studies have shown that livestock competitively affects wildlife, before this study there was virtually no direct experimental evidence that wild herbivores adversely affect livestock in Africa.

Bite RateBite Step

B) Use by several larger wildlife species was reduced in plots accessible to cattle.   Use by zebras, the most abundant wildlife species, was 48% higher in plots from which cattle are excluded than in other plots.   C) This compensation, however, was significantly affected by the presence of elephants.   In plots from which elephants had been excluded, cattle reduced the presence of zebras by 78%, but in plots accessible to elephants, cattle only reduced the presence of zebras by 22%.   This interaction occurred on all three replicate blocks, and was paralleled by a similar interaction in grass cover, suggesting that zebras are effectively tracking grass cover.   It appears the reduction in forb cover by elephants causes cattle to take fewer bites in plots to which elephants have access, resulting in higher grass cover and greater subsequent zebra use.   This interaction was unexpected, and could only have been revealed in an experiment of this crossed design.


Zebra ResearchGrass vs Dung

Ecology of Acacia drepanolobium and other woody species

We have been studying the ecology of Acacia drepanolobium , a small tree that comprises up to 95% of the woody cover in this ecosystem, in the context of the exclosure experiment.   This includes studies on its population biology, its defenses, and as part of a complementary study, its relationship with its symbiotic ants.   The largest individuals of A. drepanolobium exhibited 10% greater mortality in plots accessible to megaherbivores.   This may result eventually in significantly different densities among the experimental plots.   Long-term monitoring will be able to determine if this is indeed the case. Bush encroachment by several woody species (but not A. drepanolobium ) in the exclosures has been dramatic.   The encroachment is not due to increased recruitment, but rather to the release from suppression of small but apparently older woody plants.   In the plots open to large mammalian herbivores, individual woody plants are often only a few centimeters tall, due to intensive and repeated browsing although their stems may be up to 2cm in diameter.   When protected from large mammals, these individual plants have exhibited remarkable growth, often attaining heights of 2-4m in a few years.   Initial results suggest that each of the three major classes of herbivores suppresses a different array of woody species.   When released, these woody plant species harbor a significantly enriched avifauna.

Plant defense against mammalian herbivores

Acacia drepanolobium is eaten by a wide variety of herbivores, including large mammals, and may use a complex array of defenses.   Plant defenses may be constitutive (fixed) or induced (activated only in response to herbivore cues), and induced defenses may be relaxed following the removal of the herbivore threat. Relaxation has rarely been studied because it is a slow process requiring long-term study.   This long-term experiment is therefore ideal for testing the relaxation of induced defenses of A. drepanolobium in the absence of large mammal herbivory.   In the herbivore exclusion plots, spine length in Acacia drepanolobium has declined slowly and linearly.   After five years of mammalian herbivore exclusion, trees protected from herbivory were producing spines 40% shorter than those exposed to herbivory.   A curious exception was that those trees occupied by the ant Crematogaster nigriceps continued to produce long spines even in protected plots.   This ant species prunes it host tree , inducing the production of longer spines.   We have also found relaxations in leaf tannins, another possible defense against mammalian herbivores.   Acacias in herbivore exclosure plots have also begun to relax their investment in the food and shelter they offer to their resident symbiotic ants, which serve as an indirect defense.   After seven years of herbivore exclusion, trees have reduced the number of provisioned nectaries and swollen thorn refuges they produce for their resident ants by 25.4% and 24.8%, respectively.   These results suggest that investment in both direct and indirect defenses are expensive for A. drepanolobium and that it reduces its investment in these defenses as it experiences a reduction in the threat of herbivory. In 1998, we carried out a simulated herbivory study in the KLEE plots to examine both vegetative and defense responses of A. drepanolobium .   We caused strong compensatory vegetative responses, as measured by changes in shoot growth, but no significant changes in spine length. In 2000, more controlled simulated herbivory experiments both in the greenhouse and in the field produced significant increases in spine length.   The gradual relaxation of spine length in the absence of herbivory, contrasted with the rapid induction of greater spine length in the presence of herbivory, is striking and may be due to the differing reliability of these two signals as measures of the presence or absence of large herbivores.

Effects of mammalian exclusion on habitat heterogeneity

Savanna landscapes are often mosaics of high and low nutrient landscape features .   In particular, there exist "hot spots" of nutrient availability and plant productivity that may be in part maintained by demonstrated greater herbivore use. Our experiments seek to discern whether herbivores maintain or reduce this landscape heterogeneity, and if so, which herbivores are responsible.   Effects of large mammalian herbivores on soil chemistry did not appear during the first five years of the experiment, but have become clearer following a two-year drought.   Soils in glades, which are tree-less islands in a matrix of savanna vegetation, have greatly elevated levels of N, C, P, and K compared to soils in background vegetation.   For glades on both black cotton soils and in the adjacent red soil ecosystem, where there is strong evidence that they have their origin in old pastoral cattle exclosures.   In the black cotton soils, termite mound soils have similar chemistry and texture (and vegetation) to that in the glades.   This and other evidence suggests that the glades in this site (but on on the red soils) may be at least partly created or maintained by termite activity.   We originally predicted that large mammals were helping to maintain high soil nutrients in the glades by differentially spending time in glades as opposed to background vegetation. Pilot soil samples taken pre-treatment and after five years of mammalian herbivore removal showed no experimental effects.   However, the most recent soil sampling (2002) suggests that in the total exclosure plots, soil differences between glades and background are being accentuated, not muted, compared to control plots.    If these results are confirmed by the proposed research, it would indicate that large mammals are reducing the glade/background landscape heterogeneity, not reinforcing it.   This is likely due to the effects of cattle, the dominant herbivore, which appear to be net exporters from glades.  Kari Veblen is investigating these glades on the black cotton soils in even greater detail, particularly how they develop though time. Lauren McGoech is looking at who these spatial features interact with each other across the landscape via their edge effects.

Context-dependent interactions

In a recent review by a team of leading ecologists, Agrawal et al. outlined some of the shortcomings and opportunities for population and community ecology for the coming decade. They particularly highlighted limits to our understanding of a) “how the strength of interactions between species changes with biotic & abiotic context”, and b) the bi-directionality of interactions (feedbacks). Context-dependence and biological feedbacks at the landscape scale are fundamental themes of our work at KLEE.
Much of the early work in KLEE has demonstrated how various community and population traits respond to the effects of herbivory and rainfall (see above). Increasingly, we are also asking how interspecific interactions are contingent upon ecosystem drivers, such as herbivory, fire, and rainfall (See Table 1). For example, one particularly exciting result was that the strength of a multi-trophic cascade (herbivores-vegetation-insects-lizards) was negatively correlated with ecosystem productivity37.
Such studies require either multi-tiered experimental designs, such as manipulating competition under different herbivory regimes, or carrying out experiments over multiple years/seasons. We are carrying out both types of studies at KLEE.

Interaction Contingent upon:
Competitive suppression of zebras by cattle Elephants
Competitive suppression of cattle by wildlife Rainfall
Competition-facilitation balance among grasses Herbivory & rainfall
Competition-facilitation balance between grasses & trees Herbivory
Strength of trophic cascade (lizards) Productivity
Stability of acacia-ant mutualism Browsers
Ant community structure Browsers
Effects of herbivores on plants Rainfall
Effects of fire on woody species Elephants

Table 1: Context dependence of interactions demonstrated in the KLEE study.

KLEE collaborators

One of the great strengths of this project is that it provides a replicated, controlled experimental setting within which other scientists can carry out research. The KLEE plots have attracted numerous collaborating researchers, including projects funded by eight NSF grants:

  • Duncan Kimuyu (Karatina University): Effects of livestock and fire on wildlife species.
  • Wilfred Odadi (Egerton University): effects of wildlife and habitat heterogeneity (fire, termites mounds, glades) on cattle behaviour and production.
  • Kari Veblen (Utah State University): The ecology and management of anthropogenic landscape mosaics.
  • Amy Wolf (University of Texas): Effects of large mammalian herbivores on ecosystem geochemistry. (planned NutNet sub-experiments).
  • Lauren Hallett (University of Oregon): Mechanisms of stability and resiliences.
  • Harry Wells (Leeds University and Lolldaiga Conservancy): Investigating the impacts of livestock grazing pressure (including bomas) on savanna vegetation and the interacting effects of the presence of different wildlife guilds.
  • Corinna Riginos (The Nature Conservancy): Grazing strategies for wildlife-livestock coexistence; consequences of tree density in savannas.
  • Felicia Keesing  (Bard College): Rodent and tick ecology.
  • Todd Palmer (University of Florida), Maureen Stanton (University of California), and Alison Brody (University of Vermont): Ecology of species coexistence and mutualism in acacia ants; ecology of savanna termites
  • Ryan Langendorf and Dan Doak and (University of Colorado): Modelling species interactions.
  • Hillary Young and Doug McCauley (University of California at Santa Cruz): The ecology of disease pathogens and vectors (rodents).
  • Judith Sitters (Vrije Universiteit Brussels): Soil, plant, and animal stoichiometry.
  • Kelly Caylor (University of California at Santa Cruz): Ecohydrological impacts of large mammals herbivory an tree-grass balance, including soil respiration.
  • Jake Goheen (University of Wyoming): Assessing the impacts of large herbivores upon Acacia reproduction and establishment.
  • Rob Pringle (Princeton University): Functional role of large herbivores & consequences of their extinction; Patterns of abundance in African savannas.


National Science Foundation (LTREB x 5; REU x8; ROA x1), National Geographic Society (x3), Smithsonian Institution (James Smithson Fund), U.S. Fish & Wildlife Service (African Elephant Conservation Program), University of California.